Lions and Muons: Optimization via Stochastic Frank-Wolfe

TL;DR

This paper unifies recent optimizers Lion and Muon with the classical Stochastic Frank-Wolfe method, providing convergence guarantees and extending the framework to handle heavy-tailed noise in non-convex optimization.

Contribution

It shows Lion and Muon as special cases of Stochastic Frank-Wolfe and develops robust variants for heavy-tailed noise, with theoretical guarantees and practical improvements.

Findings

Lion and Muon are special instances of Stochastic Frank-Wolfe.

New robust variants handle heavy-tailed gradient noise with strong guarantees.

Convergence to stationarity implies convergence to KKT points under norm constraints.

Abstract

Stochastic Frank-Wolfe is a classical optimization method for solving constrained optimization problems. On the other hand, recent optimizers such as Lion and Muon have gained quite significant popularity in deep learning. In this work, building on recent initiatives, we provide a unifying perspective by interpreting these seemingly disparate methods through the lens of Stochastic Frank-Wolfe. Specifically, we show that Lion and Muon with weight decay can be viewed as special instances of a Stochastic Frank-Wolfe, and we establish their convergence guarantees in terms of the Frank-Wolfe gap, a standard stationarity measure in non-convex optimization for Frank-Wolfe methods. We further find that convergence to this gap implies convergence to a KKT point of the original problem under a norm constraint for Lion and Muon. Moreover, motivated by recent empirical findings that stochastic gradients in modern machine learning tasks often exhibit heavy-tailed distributions, we extend Stochastic Frank-Wolfe to settings with heavy-tailed noise by developing two robust variants with strong theoretical guarantees that hold for general compact convex sets without the need for a large batch size, filling the gap in the literature on Stochastic Frank-Wolfe for non-convex optimization. Our contributions in the later part of this work, in turn, yield new variants of Lion and Muon, that better accommodate heavy-tailed gradient noise, thereby enhancing their practical scope.